1 | #region License Information
|
---|
2 | /* HeuristicLab
|
---|
3 | * Copyright (C) 2002-2019 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
4 | *
|
---|
5 | * This file is part of HeuristicLab.
|
---|
6 | *
|
---|
7 | * HeuristicLab is free software: you can redistribute it and/or modify
|
---|
8 | * it under the terms of the GNU General Public License as published by
|
---|
9 | * the Free Software Foundation, either version 3 of the License, or
|
---|
10 | * (at your option) any later version.
|
---|
11 | *
|
---|
12 | * HeuristicLab is distributed in the hope that it will be useful,
|
---|
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
15 | * GNU General Public License for more details.
|
---|
16 | *
|
---|
17 | * You should have received a copy of the GNU General Public License
|
---|
18 | * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
|
---|
19 | */
|
---|
20 | #endregion
|
---|
21 |
|
---|
22 | using System;
|
---|
23 | using System.Linq;
|
---|
24 | using HeuristicLab.Algorithms.GradientDescent;
|
---|
25 | using HeuristicLab.Common;
|
---|
26 | using HeuristicLab.Core;
|
---|
27 | using HeuristicLab.Data;
|
---|
28 | using HeuristicLab.Operators;
|
---|
29 | using HeuristicLab.Optimization;
|
---|
30 | using HeuristicLab.Parameters;
|
---|
31 | using HEAL.Attic;
|
---|
32 | using HeuristicLab.PluginInfrastructure;
|
---|
33 | using HeuristicLab.Problems.DataAnalysis;
|
---|
34 | using HeuristicLab.Random;
|
---|
35 |
|
---|
36 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
37 | /// <summary>
|
---|
38 | /// Neighborhood Components Analysis
|
---|
39 | /// </summary>
|
---|
40 | [Item("Neighborhood Components Analysis (NCA)", @"Implementation of Neighborhood Components Analysis
|
---|
41 | based on the description of J. Goldberger, S. Roweis, G. Hinton, R. Salakhutdinov. 2005.
|
---|
42 | Neighbourhood Component Analysis. Advances in Neural Information Processing Systems, 17. pp. 513-520
|
---|
43 | with additional regularizations described in Z. Yang, J. Laaksonen. 2007.
|
---|
44 | Regularized Neighborhood Component Analysis. Lecture Notes in Computer Science, 4522. pp. 253-262.")]
|
---|
45 | [Creatable(CreatableAttribute.Categories.DataAnalysisClassification, Priority = 170)]
|
---|
46 | [StorableType("3C489916-B35E-408D-B225-CAA41B156065")]
|
---|
47 | public sealed class NcaAlgorithm : EngineAlgorithm {
|
---|
48 | #region Parameter Names
|
---|
49 | private const string SeedParameterName = "Seed";
|
---|
50 | private const string SetSeedRandomlyParameterName = "SetSeedRandomly";
|
---|
51 | private const string KParameterName = "K";
|
---|
52 | private const string DimensionsParameterName = "Dimensions";
|
---|
53 | private const string InitializationParameterName = "Initialization";
|
---|
54 | private const string NeighborSamplesParameterName = "NeighborSamples";
|
---|
55 | private const string IterationsParameterName = "Iterations";
|
---|
56 | private const string RegularizationParameterName = "Regularization";
|
---|
57 | private const string NcaModelCreatorParameterName = "NcaModelCreator";
|
---|
58 | private const string NcaSolutionCreatorParameterName = "NcaSolutionCreator";
|
---|
59 | private const string ApproximateGradientsParameterName = "ApproximateGradients";
|
---|
60 | private const string NcaMatrixParameterName = "NcaMatrix";
|
---|
61 | private const string NcaMatrixGradientsParameterName = "NcaMatrixGradients";
|
---|
62 | private const string QualityParameterName = "Quality";
|
---|
63 | #endregion
|
---|
64 |
|
---|
65 | public override Type ProblemType { get { return typeof(IClassificationProblem); } }
|
---|
66 | public new IClassificationProblem Problem {
|
---|
67 | get { return (IClassificationProblem)base.Problem; }
|
---|
68 | set { base.Problem = value; }
|
---|
69 | }
|
---|
70 |
|
---|
71 | #region Parameter Properties
|
---|
72 | public IValueParameter<IntValue> SeedParameter {
|
---|
73 | get { return (IValueParameter<IntValue>)Parameters[SeedParameterName]; }
|
---|
74 | }
|
---|
75 | public IValueParameter<BoolValue> SetSeedRandomlyParameter {
|
---|
76 | get { return (IValueParameter<BoolValue>)Parameters[SetSeedRandomlyParameterName]; }
|
---|
77 | }
|
---|
78 | public IFixedValueParameter<IntValue> KParameter {
|
---|
79 | get { return (IFixedValueParameter<IntValue>)Parameters[KParameterName]; }
|
---|
80 | }
|
---|
81 | public IFixedValueParameter<IntValue> DimensionsParameter {
|
---|
82 | get { return (IFixedValueParameter<IntValue>)Parameters[DimensionsParameterName]; }
|
---|
83 | }
|
---|
84 | public IConstrainedValueParameter<INcaInitializer> InitializationParameter {
|
---|
85 | get { return (IConstrainedValueParameter<INcaInitializer>)Parameters[InitializationParameterName]; }
|
---|
86 | }
|
---|
87 | public IFixedValueParameter<IntValue> NeighborSamplesParameter {
|
---|
88 | get { return (IFixedValueParameter<IntValue>)Parameters[NeighborSamplesParameterName]; }
|
---|
89 | }
|
---|
90 | public IFixedValueParameter<IntValue> IterationsParameter {
|
---|
91 | get { return (IFixedValueParameter<IntValue>)Parameters[IterationsParameterName]; }
|
---|
92 | }
|
---|
93 | public IFixedValueParameter<DoubleValue> RegularizationParameter {
|
---|
94 | get { return (IFixedValueParameter<DoubleValue>)Parameters[RegularizationParameterName]; }
|
---|
95 | }
|
---|
96 | public IValueParameter<BoolValue> ApproximateGradientsParameter {
|
---|
97 | get { return (IValueParameter<BoolValue>)Parameters[ApproximateGradientsParameterName]; }
|
---|
98 | }
|
---|
99 | public IValueParameter<INcaModelCreator> NcaModelCreatorParameter {
|
---|
100 | get { return (IValueParameter<INcaModelCreator>)Parameters[NcaModelCreatorParameterName]; }
|
---|
101 | }
|
---|
102 | public IValueParameter<INcaSolutionCreator> NcaSolutionCreatorParameter {
|
---|
103 | get { return (IValueParameter<INcaSolutionCreator>)Parameters[NcaSolutionCreatorParameterName]; }
|
---|
104 | }
|
---|
105 | #endregion
|
---|
106 |
|
---|
107 | #region Properties
|
---|
108 | public int Seed {
|
---|
109 | get { return SeedParameter.Value.Value; }
|
---|
110 | set { SeedParameter.Value.Value = value; }
|
---|
111 | }
|
---|
112 | public bool SetSeedRandomly {
|
---|
113 | get { return SetSeedRandomlyParameter.Value.Value; }
|
---|
114 | set { SetSeedRandomlyParameter.Value.Value = value; }
|
---|
115 | }
|
---|
116 | public int K {
|
---|
117 | get { return KParameter.Value.Value; }
|
---|
118 | set { KParameter.Value.Value = value; }
|
---|
119 | }
|
---|
120 | public int Dimensions {
|
---|
121 | get { return DimensionsParameter.Value.Value; }
|
---|
122 | set { DimensionsParameter.Value.Value = value; }
|
---|
123 | }
|
---|
124 | public int NeighborSamples {
|
---|
125 | get { return NeighborSamplesParameter.Value.Value; }
|
---|
126 | set { NeighborSamplesParameter.Value.Value = value; }
|
---|
127 | }
|
---|
128 | public int Iterations {
|
---|
129 | get { return IterationsParameter.Value.Value; }
|
---|
130 | set { IterationsParameter.Value.Value = value; }
|
---|
131 | }
|
---|
132 | public double Regularization {
|
---|
133 | get { return RegularizationParameter.Value.Value; }
|
---|
134 | set { RegularizationParameter.Value.Value = value; }
|
---|
135 | }
|
---|
136 | public INcaModelCreator NcaModelCreator {
|
---|
137 | get { return NcaModelCreatorParameter.Value; }
|
---|
138 | set { NcaModelCreatorParameter.Value = value; }
|
---|
139 | }
|
---|
140 | public INcaSolutionCreator NcaSolutionCreator {
|
---|
141 | get { return NcaSolutionCreatorParameter.Value; }
|
---|
142 | set { NcaSolutionCreatorParameter.Value = value; }
|
---|
143 | }
|
---|
144 | #endregion
|
---|
145 |
|
---|
146 | [StorableConstructor]
|
---|
147 | private NcaAlgorithm(StorableConstructorFlag _) : base(_) { }
|
---|
148 | private NcaAlgorithm(NcaAlgorithm original, Cloner cloner) : base(original, cloner) { }
|
---|
149 | public NcaAlgorithm()
|
---|
150 | : base() {
|
---|
151 | Parameters.Add(new ValueParameter<IntValue>(SeedParameterName, "The seed of the random number generator.", new IntValue(0)));
|
---|
152 | Parameters.Add(new ValueParameter<BoolValue>(SetSeedRandomlyParameterName, "A boolean flag that indicates whether the seed should be randomly reset each time the algorithm is run.", new BoolValue(true)));
|
---|
153 | Parameters.Add(new FixedValueParameter<IntValue>(KParameterName, "The K for the nearest neighbor.", new IntValue(3)));
|
---|
154 | Parameters.Add(new FixedValueParameter<IntValue>(DimensionsParameterName, "The number of dimensions that NCA should reduce the data to.", new IntValue(2)));
|
---|
155 | Parameters.Add(new ConstrainedValueParameter<INcaInitializer>(InitializationParameterName, "Which method should be used to initialize the matrix. Typically LDA (linear discriminant analysis) should provide a good estimate."));
|
---|
156 | Parameters.Add(new FixedValueParameter<IntValue>(NeighborSamplesParameterName, "How many of the neighbors should be sampled in order to speed up the calculation. This should be at least the value of k and at most the number of training instances minus one will be used.", new IntValue(60)));
|
---|
157 | Parameters.Add(new FixedValueParameter<IntValue>(IterationsParameterName, "How many iterations the conjugate gradient (CG) method should be allowed to perform. The method might still terminate earlier if a local optima has already been reached.", new IntValue(50)));
|
---|
158 | Parameters.Add(new FixedValueParameter<DoubleValue>(RegularizationParameterName, "A non-negative paramter which can be set to increase generalization and avoid overfitting. If set to 0 the algorithm is similar to NCA as proposed by Goldberger et al.", new DoubleValue(0)));
|
---|
159 | Parameters.Add(new ValueParameter<INcaModelCreator>(NcaModelCreatorParameterName, "Creates an NCA model out of the matrix.", new NcaModelCreator()));
|
---|
160 | Parameters.Add(new ValueParameter<INcaSolutionCreator>(NcaSolutionCreatorParameterName, "Creates an NCA solution given a model and some data.", new NcaSolutionCreator()));
|
---|
161 | Parameters.Add(new ValueParameter<BoolValue>(ApproximateGradientsParameterName, "True if the gradient should be approximated otherwise they are computed exactly.", new BoolValue()));
|
---|
162 |
|
---|
163 | NcaSolutionCreatorParameter.Hidden = true;
|
---|
164 | ApproximateGradientsParameter.Hidden = true;
|
---|
165 |
|
---|
166 | INcaInitializer defaultInitializer = null;
|
---|
167 | foreach (var initializer in ApplicationManager.Manager.GetInstances<INcaInitializer>().OrderBy(x => x.ItemName)) {
|
---|
168 | if (initializer is LdaInitializer) defaultInitializer = initializer;
|
---|
169 | InitializationParameter.ValidValues.Add(initializer);
|
---|
170 | }
|
---|
171 | if (defaultInitializer != null) InitializationParameter.Value = defaultInitializer;
|
---|
172 |
|
---|
173 | var randomCreator = new RandomCreator();
|
---|
174 | var ncaInitializer = new Placeholder();
|
---|
175 | var bfgsInitializer = new LbfgsInitializer();
|
---|
176 | var makeStep = new LbfgsMakeStep();
|
---|
177 | var branch = new ConditionalBranch();
|
---|
178 | var gradientCalculator = new NcaGradientCalculator();
|
---|
179 | var modelCreator = new Placeholder();
|
---|
180 | var updateResults = new LbfgsUpdateResults();
|
---|
181 | var analyzer = new LbfgsAnalyzer();
|
---|
182 | var finalModelCreator = new Placeholder();
|
---|
183 | var finalAnalyzer = new LbfgsAnalyzer();
|
---|
184 | var solutionCreator = new Placeholder();
|
---|
185 |
|
---|
186 | OperatorGraph.InitialOperator = randomCreator;
|
---|
187 | randomCreator.SeedParameter.ActualName = SeedParameterName;
|
---|
188 | randomCreator.SeedParameter.Value = null;
|
---|
189 | randomCreator.SetSeedRandomlyParameter.ActualName = SetSeedRandomlyParameterName;
|
---|
190 | randomCreator.SetSeedRandomlyParameter.Value = null;
|
---|
191 | randomCreator.Successor = ncaInitializer;
|
---|
192 |
|
---|
193 | ncaInitializer.Name = "(NcaInitializer)";
|
---|
194 | ncaInitializer.OperatorParameter.ActualName = InitializationParameterName;
|
---|
195 | ncaInitializer.Successor = bfgsInitializer;
|
---|
196 |
|
---|
197 | bfgsInitializer.IterationsParameter.ActualName = IterationsParameterName;
|
---|
198 | bfgsInitializer.PointParameter.ActualName = NcaMatrixParameterName;
|
---|
199 | bfgsInitializer.ApproximateGradientsParameter.ActualName = ApproximateGradientsParameterName;
|
---|
200 | bfgsInitializer.Successor = makeStep;
|
---|
201 |
|
---|
202 | makeStep.StateParameter.ActualName = bfgsInitializer.StateParameter.Name;
|
---|
203 | makeStep.PointParameter.ActualName = NcaMatrixParameterName;
|
---|
204 | makeStep.Successor = branch;
|
---|
205 |
|
---|
206 | branch.ConditionParameter.ActualName = makeStep.TerminationCriterionParameter.Name;
|
---|
207 | branch.FalseBranch = gradientCalculator;
|
---|
208 | branch.TrueBranch = finalModelCreator;
|
---|
209 |
|
---|
210 | gradientCalculator.Successor = modelCreator;
|
---|
211 |
|
---|
212 | modelCreator.OperatorParameter.ActualName = NcaModelCreatorParameterName;
|
---|
213 | modelCreator.Successor = updateResults;
|
---|
214 |
|
---|
215 | updateResults.StateParameter.ActualName = bfgsInitializer.StateParameter.Name;
|
---|
216 | updateResults.QualityParameter.ActualName = QualityParameterName;
|
---|
217 | updateResults.QualityGradientsParameter.ActualName = NcaMatrixGradientsParameterName;
|
---|
218 | updateResults.ApproximateGradientsParameter.ActualName = ApproximateGradientsParameterName;
|
---|
219 | updateResults.Successor = analyzer;
|
---|
220 |
|
---|
221 | analyzer.QualityParameter.ActualName = QualityParameterName;
|
---|
222 | analyzer.PointParameter.ActualName = NcaMatrixParameterName;
|
---|
223 | analyzer.QualityGradientsParameter.ActualName = NcaMatrixGradientsParameterName;
|
---|
224 | analyzer.StateParameter.ActualName = bfgsInitializer.StateParameter.Name;
|
---|
225 | analyzer.PointsTableParameter.ActualName = "Matrix table";
|
---|
226 | analyzer.QualityGradientsTableParameter.ActualName = "Gradients table";
|
---|
227 | analyzer.QualitiesTableParameter.ActualName = "Qualities";
|
---|
228 | analyzer.Successor = makeStep;
|
---|
229 |
|
---|
230 | finalModelCreator.OperatorParameter.ActualName = NcaModelCreatorParameterName;
|
---|
231 | finalModelCreator.Successor = finalAnalyzer;
|
---|
232 |
|
---|
233 | finalAnalyzer.QualityParameter.ActualName = QualityParameterName;
|
---|
234 | finalAnalyzer.PointParameter.ActualName = NcaMatrixParameterName;
|
---|
235 | finalAnalyzer.QualityGradientsParameter.ActualName = NcaMatrixGradientsParameterName;
|
---|
236 | finalAnalyzer.PointsTableParameter.ActualName = analyzer.PointsTableParameter.ActualName;
|
---|
237 | finalAnalyzer.QualityGradientsTableParameter.ActualName = analyzer.QualityGradientsTableParameter.ActualName;
|
---|
238 | finalAnalyzer.QualitiesTableParameter.ActualName = analyzer.QualitiesTableParameter.ActualName;
|
---|
239 | finalAnalyzer.Successor = solutionCreator;
|
---|
240 |
|
---|
241 | solutionCreator.OperatorParameter.ActualName = NcaSolutionCreatorParameterName;
|
---|
242 |
|
---|
243 | Problem = new ClassificationProblem();
|
---|
244 | }
|
---|
245 |
|
---|
246 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
247 | return new NcaAlgorithm(this, cloner);
|
---|
248 | }
|
---|
249 |
|
---|
250 | public override void Prepare() {
|
---|
251 | if (Problem != null) base.Prepare();
|
---|
252 | }
|
---|
253 | }
|
---|
254 | }
|
---|